Self-Attention Deep Learning for Medical Image Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Medical imaging technologies like PET face challenges with image quality due to factors such as radiation dose reduction, shortened scan times, and patient movement, leading to noise and artifacts that degrade image quality.
Innovation Solution
The implementation of a deep learning-based system that uses a self-attention mechanism and adaptive deep learning framework to enhance medical image quality by generating attention feature maps and improving image reconstruction from low-count projection data without modifying existing imaging systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If radiation dose is reduced, then patient safety is improved, but image quality deteriorates due to increased noise
Solution Approach 1:
The patent uses deep learning models to create a virtual copy of the imaging process. The neural network learns the mapping between low-dose/noisy images and high-quality images by analyzing training data, effectively copying the quality-enhancing transformation without requiring physical changes to the imaging hardware or increasing radiation exposure.
Solution Approach 2:
The patent replaces traditional mechanical/image-processing approaches with artificial intelligence-based deep learning models. Instead of using conventional noise reduction algorithms that require complex processing pipelines, the system uses neural networks to directly transform low-quality images into high-quality images, substituting mechanical processing with intelligent computation.
2Loss of time
If scan time is shortened, then patient comfort is improved, but image quality deteriorates due to insufficient data collection
Solution Approach 1:
The patent applies preliminary action by pre-training deep learning models on extensive datasets of high-quality medical images before actual scanning. The model learns the relationship between scan data and image quality in advance, enabling it to reconstruct high-quality images from short, low-quality scans without requiring additional data collection time during the actual imaging procedure.
Solution Approach 2:
The system creates a computational copy that simulates the imaging process and learns to reconstruct images from incomplete or low-quality data. By copying the characteristics of high-quality images through neural network training, the system can generate accurate reconstructions from accelerated scans that would traditionally be too brief for adequate data collection.
3Device complexity
If conventional image processing is used, then hardware requirements are low, but image enhancement capability is insufficient
Solution Approach 1:
The patent substitutes traditional hardware-based image processing with software-based deep learning models. Instead of requiring complex hardware systems for noise reduction and image enhancement, the system uses neural networks that can be implemented through software, replacing mechanical processing capabilities with intelligent computation that achieves superior enhancement results.
Solution Approach 2:
The patent changes the fundamental parameters of image processing by transitioning from linear, fixed-parameter algorithms to non-linear, learnable transformations. The deep learning models adapt their parameters during training to optimize image enhancement, allowing the system to achieve superior capability without increasing hardware complexity by leveraging computational intelligence rather than computational power alone.
Data Source
AI summary
A computer-implemented method is provided for improving image quality. The method comprises: acquiring, using a medical imaging apparatus, a medical image of a subject, wherein the medical image is acquired with shortened scanning time or reduced amount of tracer dose; applying a deep learning network model to the medical image to generate one or more feature attention maps a medical image of the subject with improved image quality for analysis by a physician.